基于GPR-LSTM的地震热红外背景场的构建方法

doi:10.3969/j.issn.0253-4967.2024.02.015

摘要/Abstract

摘要：

地震监测是一项非常重要且具有挑战性的任务, 遥感技术的不断发展加强了在宏观尺度上对地球表面的监测能力。研究表明, 地震前通常都会出现地表温度异常升高的现象, 因此各种异常提取算法被应用于地震热异常研究中。其中, 基于背景场的提取方法由于具有较强的机理解释性而受到广泛应用。然而, 以往基于背景场的异常提取方法更多将背景场限定于某一固定阈值, 忽略了受外界因素(非震)影响导致的地表温度的小范围正常波动。据此, 文中提出了一种基于GPR-LSTM的地震热红外背景场的构建方法。该方法包括两大部分: 震期年变基准场的建立、实际LST的残差波动范围计算及背景场的构建。基于MODIS地表温度产品, 以2008年四川汶川和新疆于田地震为研究对象, 使用所述方法对地震前兆热异常信息进行提取与分析, 经过实验得出以下结论: 1)地震热异常通常沿青藏高原的断层分布, 这不仅证明了文中方法能够减弱地表温度数据中噪声的干扰, 同时也证明该方法在热异常信息提取方面的有效性; 2)地震年份的构造活动比非地震年份更加活跃, 导致地表温度的异常增温更加明显; 3)不同地震案例震前的热异常时空特征各不相同。

关键词: 背景场, 热异常, 地震前兆, GPR, LSTM

Abstract:

Seismic monitoring is a very important and challenging task. The continuous development of remote sensing technology has strengthened our ability to monitor the Earth’s surface on a macro scale. Research shows that an abnormal rise in surface temperature usually occurs before an earthquake, so a variety of anomaly extraction algorithms have been applied to the study of seismic thermal anomalies. Among them, the extraction method based on background field is widely used because of its strong mechanism interpretation. However, the previous anomaly methods based on the background field mainly limit the background field to a certain fixed threshold value, and ignore the small range of normal LST fluctuations caused by external factors(non-seismic). Therefore, a method of constructing a seismothermal infrared background field based on GPR-LSTM is proposed in this paper. The main idea of the method is that the LST background field is obtained by adding the established annual variable reference field and the fluctuation range of normal LST. First, the LST exhibits a range of fluctuations due to non-seismic factors such as solar radiation, weather, and human activities. Therefore, it is not reasonable to take a fixed value of the LST background field but it should have a certain fluctuation range. Therefore, the dynamic fluctuation characteristics of the background field should be reflected in the construction process of the background field in this study. Secondly, the reason why this method uses the LSTM model to predict the annual variable reference field of the earthquake period based on the annual variable reference field of the non-earthquake period is that on the one hand, the LSTM model can predict the law of long time series data, so it can better learn the annual variation law of LST in the non-earthquake year. At the same time, the LST trend of increasing or decreasing year by year caused by climate change can be predicted, which is conducive to more truly describing the LST trend of the real background field in the year of the earthquake period.

The method includes two parts: the establishment of the annual variable reference field, the calculation of the residual fluctuation range of the actual LST and the construction of the background field. Based on the MODIS surface temperature product, the precursory thermal anomaly information of the 2008 Wenchuan earthquake in Sichuan Province and the Yutian earthquake in Xinjiang Province was extracted and analyzed by using the proposed method. Based on the MODIS surface temperature data of the 2008 Wenchuan earthquake in Sichuan Province and the Yutian earthquake in Xinjiang Province, the proposed method is used to detect and analyze the thermal anomalies before the earthquake. The following conclusions are drawn:

(1)The established LST background field on the Qinghai-Xizang Plateau is consistent with the actual variation law of LST and is relatively stable on the whole. In the time dimension, the background field of LST shows the annual variation of high summer and low winter. In terms of spatial dimension, the established LST background field is generally high in the south, low in the middle and low in the north, and the temperature value of the background field is the lowest at the highest altitude of the Qinghai-Xizang Plateau.

(2)The thermal anomalies obtained based on the new algorithm are usually distributed along the fault zone of the Qinghai-Xizang Plateau, and the anomaly evolution law is obvious. For the Yutian earthquake in Xinjiang, the trend of thermal anomalies is consistent with that of the fault zone in most cases, and the thermal anomalies are mainly distributed in the northern margin of the fault zone. For the Wenchuan earthquake in Sichuan Province, with the onset time approaching, the thermal anomalies gradually moved southward along the Longmenshan fault zone from the north margin, filled the entire fault zone before the earthquake, and disappeared after the earthquake. This proves the validity of the proposed background field construction method.

(3)Compared with non-seismic years, the spatial characteristics of thermal anomalies along faults are more obvious in seismic years, and the duration and amplitude of the anomalies are longer. This indicates that the tectonic activity in seismic years is more active than that in non-seismic years, resulting in more significant abnormal warming of surface temperature.

(4)The evolution law of thermal anomalies of earthquakes with magnitude 7 or above on the Qinghai-Tibet Plateau can be summarized as incubation—disappearance—accumulation—disappearance—earthquake occurrence, and the strength of thermal anomalies appears in the way of repeated changes. This indicates that the thermal anomaly is not continuously enhanced or weakened before the earthquake but shows the characteristics of repeated fluctuations between strong and weak until the earthquake eruption thermal anomaly disappears.

In conclusion, the proposed method can not only ensure that the established background field is not affected by tectonic factors but also fully consider the small range of normal fluctuation of surface temperature caused by non-tectonic factors, which makes the extraction of thermal infrared anomaly information more accurate. This provides new technical means and ideas for the extraction of earthquake anomaly information and is helpful for the further development of earthquake monitoring.

Key words: Background field, Anomaly of heat, Precursor of earthquake, GPR, LSTM

宋冬梅, 张曼玉, 单新建, 崔建勇, 王斌. 基于GPR-LSTM的地震热红外背景场的构建方法[J]. 地震地质, 2024, 46(2): 492-511.

SONG Dong-mei, ZHANG Man-yu, SHAN Xin-jian, CUI Jian-yong, WANG Bin. CONSTRUCTION METHOD OF SEISMOTHERMAL INFRARED BACKGROUND FIELD BASED ON GPR-LSTM[J]. SEISMOLOGY AND GEOLOGY, 2024, 46(2): 492-511.

图/表 15

图1 研究区示意图 2个紫色边框所示区域分别代表2个不同地震案例的子研究区: 1 新疆于田地震; 2 四川汶川地震。图中白色部分表示地表温度数据缺值

Fig. 1 Schematic diagram of research area.

图2 本文所述方法的技术流程图

Fig. 2 A technical flow chart of the proposed method.

图3 GPR曲线拟合模型示意图

Fig. 3 Diagram of GPR curve fitting model.

图4 LSTM神经网络结构示意图

Fig. 4 Schematic diagram of LSTM neural network structure.

图5 非震期残差示意图图中横坐标代表残差值x的数量(即图中1个点代表1个时刻)

Fig. 5 Diagram of residuals during non-seismic periods.

图6 2008年四川汶川不同阈值范围热异常分布图

Fig. 6 Thermal anomaly distribution map of different threshold ranges in Wenchuan, Sichuan, 2008.

图7 青藏高原地区的地表温度背景场(2003—2007年)

Fig. 7 Background field of surface temperature over the Qinghai-Xizang Plateau(2003—2007).

图8 2008年新疆于田热异常分布图

Fig. 8 Distribution map of Yutian thermal anomaly in Xinjiang, 2008.

图9 2008年四川汶川热异常分布图

Fig. 9 Thermal anomaly distribution map in Wenchuan, Sichuan, 2008.

图10 2004年新疆于田热异常分布图

Fig. 10 Distribution map of Yutian thermal anomaly in Xinjiang, 2004.

图11 2006年四川汶川热异常分布图

Fig. 11 Thermal anomaly distribution map in Wenchuan, Sichuan, 2006.

图12 新疆于田震前震后的异常像元数量

Fig. 12 Anomalous pixel quantity before and after the Yutian earthquake in Xinjiang.

图13 四川汶川震前震后的异常像元数量

Fig. 13 Anomalous pixel quantity before and after the Wenchuan earthquake in Sichuan.

图14 2008年新疆于田地区的异常判定

Fig. 14 Abnormal judgment of Yutian area at Xinjiang in 2008.

图15 2008年四川汶川地区的异常判定

Fig. 15 Anomaly determination in Wenchuan, Sichua in 2008.

参考文献 30

[1]	蔡丽雯, 李戈云. 2014. 2013年甘肃岷县漳县6.6级地震前热红外异常分析[J]. 高原地震, 26(4): 5—9.
	CAI Li-wen, LI Ge-yun. 2014. Analysis on the thermal infrared anomaly before Minxian-Zhangxian M_S6.6 in 2013 in Gansu Province[J]. Plateau Earthquake Research, 26(4): 5—9(in Chinese).
[2]	陈梅花, 邓志辉, 马晓静, 等. 2007. 断裂带内外温差法在震前红外异常研究中的应用[J]. 地震地质, 29(4): 863—872.
	CEHN Mei-hua, DENG Zhi-hui, MA Xiao-jing, et al. 2007. Application of the inside-outside temperature relation analysis method in study on satellite infrared anomalies prior to earthquake[J]. Seismology and Geology, 29(4): 863—872(in Chinese).
[3]	陈顺云. 2006. 提取现今构造活动信息的定量热红外遥感方法研究[D]. 北京: 中国地震局地质研究所.
	CHEN Shun-yun. 2006. A study on the quantitative thermal infrared remote sensing method for extracting information of current tectonic activity[D]. Institute of Geology, China Earthquake Administration, Beijing(in Chinese).
[4]	陈顺云, 刘力强, 陈国强, 等. 2004. 岩石红外比辐射率简易测定方法及意义[J]. 地震地质, 6(3): 512—518. doi: 10.3969/j.issn.0253-4967.2004.03.015.
	CHEN Shun-yun, LIU Li-qiang, CHEN Guo-qiang, et al. 2004. A simple method for measuring emissivity of rock and its significance[J]. Seismology and Geology, 6(3): 512—518(in Chinese).
[5]	黄广思. 1993. 地温遥感预报地震的原理和方法[J]. 地壳形变与地震, 21(1): 23—29.
	HUANG Guang-si. 1993. Principle and methods of earthquake prediction by temperature remote sensing[J]. Crustal Deformation and Earthquake, 21(1): 23—29(in Chinese).
[6]	康春丽, 张艳梅, 刘德富, 等. 2009. 汶川8.0级大地震的长波辐射征象[J]. 地震, 29(1): 116—120.
	KANG Chun-li, ZHANG Yan-mei, LIU De-fu, et al. 2009. Long-wave-radiation patterns prior to the Wenchuan M8.0 earthquake[J]. Earthquake, 29(1): 116—120(in Chinese).
[7]	赁常恭, 王宣吉, 强祖基. 1992. 卫星热红外异常增温与常熟临震前兆[J]. 科学通报, 24(9): 821—824.
	LIN Chang-gong, WANG Xuan-ji, QIANG Zu-ji. 1992. Abnormal thermal infrared warming of satellite and precursory of Changshu earthquake[J]. Chinese Science Bulletin, 24(9): 821—824(in Chinese).
[8]	刘德富, 罗灼礼, 彭克银. 1997. 强烈地震前的OLR异常现象[J]. 地震, 4(2): 126—132.
	LIU De-fu, LUO Zhuo-li, PENG Ke-yin. 1997. OLR anomalies before strong earthquakes[J]. Earthquakes, 4(2): 126—132(in Chinese).
[9]	刘雪梅. 2014. 基于热异常信息的神经网络地震预测初探[D]. 青岛: 中国石油大学(华东).
	LIU Xue-mei. 2014. The neural network prediction based on the thermal anomaly information[D]. China University of Petroleum(East China), Qingdao(in Chinese).
[10]	孟亚飞, 孟庆岩, 周世健, 等. 2020. 基于卫星数据稳健分析技术统计分析新疆地区热异常与地震的关系[J]. 地震学报, 42(2): 205—215, 245.
	MENG Ya-fei, MENG Qing-yan, ZHOU Shi-jian, et al. 2020. The relationship between thermal anomaly and earthquake in Xinjiang region is analyzed based on robust analysis of satellite data[J]. Acta Seismologica Sinica, 42(2): 205—215, 245(in Chinese).
[11]	强祖基, 徐秀登, 赁常恭. 1990. 卫星热红外异常: 临震前兆[J]. 科学通报, 6(17): 1324—1327.
	QIANG Zu-ji, XU Xiu-deng, LIN Chang-gong. 1990. Satellite thermal infrared anomaly: Precursor of impending earthquake[J]. Chinese Science Bulletin, 6(17): 1324—1327(in Chinese).
[12]	沈赟. 2009. 基于高斯过程的时间序列分析[D]. 上海: 上海交通大学.
	SHEN Yun. 2009. Time series analysis based on Gaussian process[D]. Shanghai Jiao Tong University, Shanghai(in Chinese).
[13]	石欣雪. 2022. 基于微波亮温数据的地震异常提取判定与分析[D]. 鞍山: 辽宁科技大学.
	SHI Xin-xue. 2022. Extraction judgment and analysis of seismic anomalies based on microwave brightness temperature data[D]. University of Science and Technology Liaoning, Anshan(in Chinese).
[14]	宋冬梅, 臧琳, 单新建, 等. 2016. 基于LST年趋势背景场的地震热异常提取算法[J]. 地震地质, 38(3): 680—695. doi: 10.3969/j.issn.0253-4967.2016.03.014.
	SONG Dong-mei, ZANG Lin, SHAN Xin-jian, et al. 2016. A study on the algorithm for extracting earthquake thermal infrared anomalies based on the yearly trend of LST[J]. Seismology and geology, 38(3): 680—695(in Chinese). DOI
[15]	孙珂, 单新建, Ouzounov D, 等. 2017. 基于多轨道卫星观测数据分析尼泊尔地震长波辐射特征[J]. 地球物理学报, 60(9): 3457—3465. DOI
	SUN Ke, SHAN Xin-jian, Ouzounov D, et al. 2017. Analyzing long wave radiation data associated with the 2015 Nepal earthquakes based on multi-orbit satellite observations[J]. Chinese Journal of Geophysics, 60(9): 3457—3465(in Chinese).
[16]	王辉, 欧阳缮, 廖可非, 等. 2021. 基于深度学习的GPR B-SCAN图像双曲线检测方法[J]. 电子学报, 49(5): 953—963. DOI
	WANG Hui, OUYANG Shan, LIAO Ke-fei, et al. 2021. GPR B-SCAN image hyperbola detection method based on deep learning[J]. Acta Electronica Sinica, 49(5): 953—963(in Chinese). DOI
[17]	王西刚. 2020. 基于高斯过程回归的时序预测算法研究与应用[D]. 杭州: 浙江理工大学.
	WANG Xi-gang. 2020. Research and application of time series prediction algorithm based on Gaussian process regression[D]. Zhejiang University of Science and Technology, Hangzhou(in Chinese).
[18]	王一博. 2014. MODIS地震热异常的数据处理与算法研究[D]. 青岛: 中国石油大学(华东).
	WANG Yi-bo. 2014. A study on the data processing and method for thermal infrared anomalies associated with earthquakes based on MODIS LST products[D]. China University of Petroleum(East China), Qingdao(in Chinese).
[19]	臧琳. 2017. 基于红外异常的发震信号识别与危险性评估[D]. 青岛: 中国石油大学(华东).
	ZANG Lin. 2017. Seismic signals identification and risk assessment based on thermal infrared anomalies[D]. China University of Petroleum(East China), Qingdao(in Chinese).
[20]	张璇, 张元生, 郭晓, 等. 2017. 尼泊尔8.1级地震卫星热红外异常解析[J]. 地学前缘, 24(2): 227—233. DOI
	ZHANG Xuan, ZHANG Yuan-sheng, GUO Xiao, et al. 2017. Analysis of thermal infrared anomaly in the Nepal M_S8.1 earthquake[J]. Earth science frontiers, 24(2): 227—233(in Chinese).
[21]	张元生, 郭晓, 钟美娇, 等. 2010. 汶川地震卫星热红外亮温变化[J]. 科学通报, 55(10): 904—910.
	ZHANG Yuan-sheng, GUO Xiao, ZHONG Mei-jiao, et al. 2010. Thermal infrared brightness temperature variation of Wenchuan earthquake satellite[J]. Chinese Science Bulletin, 55(10): 904—910(in Chinese).
[22]	朱传华, 单新建, 张国宏, 等. 2019. 汶川地震热异常与构造应力关联的数值模拟[J]. 地震地质, 41(6): 1497—1510. doi: 10.3969/j.issn.0253-4967.2019.06.012.
	ZHU Chuan-hua, SHAN Xin-jian, ZHANG Guo-hong, et al. 2019. Two dimensional model on relation between thermal anomaly before Wenchuan earthquake and tectonic stress[J]. Seismology and geology, 41(6): 1497—1510(in Chinese). DOI
[23]	Chen S, Liu P, Feng T, et al. 2020. Exploring changes in land surface temperature possibly associated with earthquake: Case of the April 2015 Nepal M_W7.9 earthquake[J]. Entropy, 22(4): 1—23.
[24]	Cui J Y, Zhang M Y, Song D M, et al. 2022. MODIS land surface temperature product reconstruction based on the SSA-BiLSTM Model[J]. Remote Sensing, 14(4): 1—16.
[25]	Hafeez A, Ehsan M, Abbas A, et al. 2022. Machine learning-based thermal anomalies detection from MODIS LST associated with the M_W7.7 Awaran, Pakistan earthquake[J]. Natural Hazards, 111(4): 1—19.
[26]	Khalili M, Eskandar S S A, Panah S K A. 2020. Thermal anomalies detection before Saravan earthquake(April 16th, 2013, M_W=7.8)using time series method, satellite, and meteorological data[J]. Journal of Earth System Science, 129(1): 1—10.
[27]	Pergola N, Aliano C, Coviello I, et al. 2010. Using RST approach and EOS-MODIS radiances for monitoring seismically active regions: A study on the 6 April 2009 Abruzzo earthquake[J]. Natural Hazards & Earth System Sciences, 10(2): 239—249.
[28]	Sekertekin A, Inyurt S, Yaprak S. 2020. Pre-seismic ionospheric anomalies and spatio-temporal analyses of MODIS land surface temperature and aerosols associated with Sept., 24, 2013 Pakistan earthquake[J]. Journal of Atmospheric and Solar-Terrestrial Physics, 20: 1—18.
[29]	Song D, Xie R, Zang L, et al. 2018. A new algorithm for the characterization of thermal infrared anomalies in tectonic activities[J]. Remote Sensing, 10(12): 1—33.
[30]	Tramutoli V, Corrado R, Filizzola C, et al. 2015. One year of RST based satellite thermal monitoring over two Italian seismic areas[J]. Bollettino di Geofisica Teorica ed Applicata, 56(2): 275—294.